Noise: A Flaw in Human Judgment

by Daniel Kahneman

the belief that it is too expensive to reduce noise is not always wrong. In short, we have to compare the benefits of noise reduction with the costs.

The Rhetoric of Reaction, economist Albert Hirschman

three common objections to reform efforts. First, such efforts might be perverse, in the sense that they will aggravate the very problem they are intended to solve. Second, they might be futile; they might not change things at all. Third, they put other important values in jeopardy (such as when an effort to protect labor unions and the right to unionize is said to hurt economic growth). Perversity, futility, and jeopardy might be offered as objections to noise reduction,

prosaic

predictors that are highly correlated with race or gender. For example, height and weight are correlated with gender, and the place where people grew up or where they live might well be correlated with race.

discrimination could also come from the source data. If an algorithm is trained on a data set that is biased,

A rule-bound system might eliminate noise, which is good, but it might also freeze existing norms and values, which is not so good.

noise-reduction efforts need not and should not be permanent. If such efforts take the form of firm rules, those who make them should be willing to make changes over time.

preposterous

Some companies and universities forbid people to engage in “wrongdoing,” without specifying what that means. The inevitable result is noise, which is not good and may even be very bad. But if there is a specific list of what counts as wrongdoing, then terrible behavior that is not explicitly covered by the list will end up being tolerated.

When we cannot easily design rules that ban all conduct that ought to be prohibited, we have a distinctive reason to tolerate noise, or so the objection goes.

In some circumstances, clear, defined rules eliminating noise do give rise to the risk of evasion. And this risk might be a reason to adopt some other strategy for reducing noise, such as aggregation, and perhaps to tolerate an approach that allows for some noise. But the words might be are crucial. We need to ask how much evasion there would be—and how much noise there would be. If there is only a little evasion and a lot of noise, then we are better off with approaches that reduce noise.

To know whether a noisy system imposes more deterrence, we need to know whether potential wrongdoers are risk-averse or risk-seeking. And if we want to increase deterrence, wouldn’t it be better to increase the penalty and eliminate the noise? Doing that would eliminate unfairness as well.

If the goal is to reduce noise or decide how and whether to do so (and to what degree), it is useful to distinguish between two ways of regulating behavior: rules and standards.

Rules are meant to eliminate discretion by those who apply them; standards are meant to grant such discretion.

Rules have an important feature: they reduce the role of judgment. On that count, at least, judges (understood to include all those who apply rules) have less work to do. They follow the rules. For better or worse, they have far less room to maneuver.

Standards are altogether different. When standards are in place, judges have to do a lot of work to specify the meaning of open-ended terms. They might have to make numerous judgments to decide what counts as (for example) “reasonable” and “feasible.”

Algorithms work as rules, not standards.

whenever firms, organizations, societies, or groups are sharply divided, it might be far easier to generate standards than rules.

People might agree that a constitution should protect freedom of speech, without deciding whether it should protect commercial advertising, threats, or obscenity.

Leaders might want to come up with rules but, as a practical matter, might not be able to agree on them.

“Take a pill every morning and every night” is a rule; “take a pill whenever you feel you need it” is a standard.

a disability matrix.

bureaucratic justice.

jury nullification refers to situations in which juries simply refuse to follow the law, on the ground that it is senselessly rigid and harsh.

(1) the costs of decisions and (2) the costs of errors.

With standards, the costs of decisions can be very high

Exercising judgment can be ...

With rules, the costs of decisions are typically much lower.

Rules may be straightforward to apply once they are in place, but before a rule is put in place, someone has to decide what it is. Producing a rule can be hard. Sometimes it is prohibitively costly. Legal systems and private companies therefore often use words such as reasonable, prudent, and feasible.

Sensible organizations well understand that the amount of discretion they grant is closely connected with the level of trust they have in their agents.

“Kadi justice,” which he understood as informal, ad hoc judgments undisciplined by general rules.

“Rules simplify life, and reduce noise. But standards allow people to adjust to the particulars of the situations.”

Noise is the unwanted variability of judgments,

judgment is a form of measurement in which the instrument is a human mind.

Judgments informally integrate diverse pieces of information into an overall assessment. They are not computations, and they do not follow exact rules.

Some judgments are predictive, and some predictive judgments are verifiable; we will eventually know whether they were accurate.

But many judgments, including long-term forecasts and answers to fictitious questions, are unverifiable. The quality of such judgments can be assessed only by the quality of the thought process that produces them.

many judgments are not predictive but evaluative:

The phrase judgment call implies both the possibility of disagreement and the expectation that it will be limited.

bounded disagreement.

The mean of squared errors (MSE) has been the standard of accuracy in scientific measurement for two hundred years.

As measured by MSE, bias and noise are independent and additive sources of error.

Variability as such is unproblematic in some judgments, even welcome. Diversity of opinions is essential for generating ideas and options. Contrarian thinking is essential to innovation. A plurality of opinions among movie critics is a feature, not a bug. Disagreements among traders make markets. Strategy differences among competing start-ups enable markets to select the fittest.